On Sep 18th 2019, we joined the Azure PlayFab team for a game developer focused meetup at Microsoft's Reactor in Tel-Aviv. Following a great overview and demo by James Gwertzman (co-founder and general manager of PlayFab), we gave a presentation covering key aspects that game developers should keep in mind while designing and building their games. With these in place, it would be feasible and cost effective to implement AI-based features and capabilities in the future.
In this post, I want to share some of the key takeaways we've discussed.
Player Churn Prediction
Player Churn is a common problem in the gaming industry. In fact, even promising games, suffer from it. The most fundamental requirement to deal with player churn is the ability to collect data, store and analyze it.
Once you figure out the type of churn you are experiencing, you can use the collected data to create features that you will use to train your prediction model. These features, for churn, should cover different aspects of the player life in and around the game.
Last but not least is the ability to act on your predictions. Having great predictions for future churn is only useful if you are able to take preventive measures. This means you want to make sure that you are able to drive difference in experiences in your game in real-time based on individual player's churn prediction.
AI Agents for Game Exercising
While great visuals, awesome game-play and intriguing story are key ingredients for a successful game, lack of quality remains one of the causes of failure - recovering from first day one-star reviews in nearly impossible.
AI Agents are trained to play like humans. Like humans, they want to win and just like humans, they will game the system if they can. But unlike humans, AI Agents never get tired and they can play your game again and again, 1000's of times - much more than an average user will.
For AI Agents to be able to play your game, there are three requirements:
Observation - being able to 'see' what is going on on the game board at any given moment is essential for the agent learning process
Control - being able to control the game. The more complex your control is, the harder it is for the agent to learn to use it wisely
Rewards - being able to clearly understand whether a specific action is a good one depends on the ability to connect it to a reward. if it is hard to understand the value of an action in your game, it will be hard for the agent to master it.
Addressing these aspects makes building an AI Agent feasible but it also makes your game easier to use for human players!
In summary, AI for Gaming is an advantage you want to leverage at some point. Churn Prediction and AI Agents are just a couple of examples, the sky is the limit to the number of challenges you can tackle with the help of AI. We recommend game developers start small, but focusing on a specific, well defined challenge, but do so soon.
At Agent Factory, we have the experience and knowledge to help out - so, don't hesitate to reach out, maybe together we can introduce some AI magic into your next game.
Comments